Introducing TimesFM-3, a state-of-the-art time series foundation model that enables accurate multivariate time series forecasting in a single forward pass, significantly outperforming other forecasting models across major benchmarks. More on the blog →https://t.co/uSlnIdUJ4Q https://t.co/NfbONpFYDz

Today, we're sharing new research on Solaris, our first Interface World Model. Solaris is a new kind of operating system that generates interactive interfaces frame by frame, in real time, with no code. We find that Solaris outperforms frontier LLMs when generating new interfaces, across structural similarity and information retention. Read more and request early access at the link below.

We just killed…everything. Introducing Monid, the OpenRouter for agent tools. Today we crossed 4M agent transactions. And raised a $2.1M pre-seed to kill every subscription in the world. Your agents can now: > discover, run, and pay for tools at runtime > one connection to 1,800 APIs: SEO, leads, search, ecommerce, stocks, video/image/music/3D gen, private company data > all pay per call, 0 subscriptions

Interesting paper from Tencent. Tencent trains an agent to manage its own working context, and assigns credit at the level of individual context edits. Long-horizon tasks force a model to retrieve, integrate and maintain scattered information across many turns, and keeping every interaction history makes the working context grow without bound. Recent proactive methods let a model edit its own context with tools, but the toolset stops at search, deletion and summarization. ContextPilot adds global planning, long-term memory and adaptive soft compression, so the agent can offload information rather than only discard it. The training side is where it gets interesting. Standard RL hands the final trajectory reward to every intermediate edit equally. ContextPilot uses context and entropy variation to find which editing decisions actually mattered, samples branches at those points, and estimates action-level advantages from all branched trajectories passing through that edit. On long-context QA and deep search it beats existing baselines across several base models while holding a more compact working context. Code is available. Paper: https://t.co/buM6galu4x Chat with Paper: https://t.co/VN5J5PO4Rt

Loop engineering has emerged as a new skill for AI engineers But there is very little research measuring how effective it is. The best results on full tasks in a new benchmark is ~25%. LoopArena from AMAP evaluates the outer loop rather than the coding agent. A Controller model receives a structured summary after each round and instructs a separate fixed Worker agent on what to do or verify next, or decides to stop. Holding the Worker constant makes the result readable, since an end-to-end run cannot tell you whether success came from the guidance or from the agent carrying it out. The named failure modes will be familiar to anyone running long agent sessions: - Trusting a stale progress note - Skipping needed verification - Spending budget in the wrong direction - Stopping before the task is safe to submit Paper: https://t.co/ZGiooDsN34 Chat with Paper: https://t.co/jSeFFp3CBc

Made a Microduck sticker pack because apparently owning the robot wasn’t enough. Thoughts? https://t.co/CNJpSPggnn

Made a Microduck sticker pack because apparently owning the robot wasn’t enough. Thoughts? https://t.co/CNJpSPggnn

LoopArena Benchmarking Models as Runtime Controllers for Loop Engineering paper: https://t.co/GEzWnPfqjr https://t.co/1jqniijCDn

🚀 August brought a lot of updates to @code! Some highlights: 📄 Review Markdown diffs while keeping the document editable 🦆 Get a second opinion on agent work with /rubber-duck 🌐 Auto-reload HTML files in the Integrated Browser 🎙️ Dictate prompts in multiple languages, with speech recognition running locally ✨ Plus, more ways to organize agent sessions, navigate conversations, and work with models. 🔗 Explore the latest updates: https://t.co/oaUDjzERsh Happy coding! 💙

You’re too kind. 🙏 I don’t know how much value there is in ideas without execution but I’m happy to give it a go! Here’s one — and it’s quite meta, because it’s about what breaks when AI slop makes certain things too cheap. If anyone would like to build on this, I’d be happy to (1) be an adviser (2) take it as a signal to clean up and publish more ideas. https://t.co/ZUxswR3emX

If you missed our opening remarks: - 4th world modeling workshop: Aspen, CO, February, world models for physics, https://t.co/mrnm5cwQqC - 1st world modeling conference: Bay Area, CA, May, https://t.co/v8KSoNBKqM In collaboration with @ylecun @LambdaAPI @amilabs and more TBA! https://t.co/Q2y0R7dNbl

One of those times you feel the urgent need for palette cleansing with Drew McDermott's "Artificial Intelligence Meets Natural Stupidity" https://t.co/LGja4Ed3Ff (or this: https://t.co/6BU2ZbczEm )

NEW WORLD MODEL: @ylecun's team is back with an efficient model! This project involves @ylecun, @lukaskuhn77, @lucasmaes_, @quentinlldc, and @randall_balestr. A couple definitions first: - DINO: self-DIstillation with NO labels. A self-supervised image model (Meta, 2021) where a student network learns to match a teacher (an EMA copy of itself) across two crops of the same image, with no labels and no negatives. - SIGReg: a regularizer that prevents embedding collapse by forcing the embeddings to match an isotropic Gaussian, tested with a normality test (Epps–Pulley) on many random 1-D projections instead of in full dimension. LeVJEPA is a self-supervised video pretraining method, released with open code, weights, and checkpoints. It learns a video representation by pushing the embeddings of global and local crops of the same clip together (an invariance loss), while a regularizer called SIGReg forces the embeddings toward an isotropic Gaussian to provably prevent representation collapse. Unlike V-JEPA and V-JEPA 2 it uses a single shared encoder with a projector and no target network, no predictor and no stop-gradient. It drops 95% of tokens per view, uses block-causal attention (each frame attends only to past frames), and has a single loss weight. It is evaluated purely as a representation learner via frozen probing on ImageNet-1K, Something-Something-v2 and Kinetics-400, not on any robot. What I find interesting, is that V-JEPA and V-JEPA 2 need an EMA target encoder, stop-gradients and a capacity-limited predictor to avoid collapse; LeVJEPA drops all of it for one shared encoder plus projector, preventing collapse instead with the SIGReg regularizer under a provable guarantee and a single hyperparameter. The "P" (predictor) in JEPA is effectively gone. LeVJEPA is also less compute intensive: - 5.6x to 20.8x lower total pretraining compute than V-JEPA 2 - 7.6 points higher on ImageNet-1K at matched FLOPs - trains at batch size 128 within 8GB where V-JEPA 2 saturates at batch size 28 Also worth mentioning: ImageNet-1K accuracy rises monotonically with the token-drop rate, from 33.9% at rho = 0 to 47.6% at rho = 0.95. The aggressive dropping is actually doing regularization work. On the JEPA-versus-DINO debate: - it loses to DINOv2 by 3.1 points on ImageNet-1K (appearance, static) - but wins on Something-Something-v2 by nearly 2x (motion, temporal) - and beats V-JEPA 2 by 1.9 points on ViT-L at 5.6x lower cost. -> optimized for temporal and motion understanding per compute dollar.

Actual random reply from @grok. Its humor is exquisite. At least, I hope it’s humor 😜 https://t.co/XnuOun4Cnc

@YahooNewsTopics https://t.co/WjOuJgtVIn

@shields_pikes https://t.co/WjOuJgtVIn

@IsoMitsuo https://t.co/WjOuJgtVIn

@goofyninjaaa https://t.co/WjOuJgtVIn

@suni_code https://t.co/WjOuJgtVIn

@MITSloan https://t.co/eZ6wdBTGvE

@loop_clips https://t.co/rblR72X0K9

@DukeDain1123 https://t.co/rblR72X0K9

@I_xco11 https://t.co/rblR72X0K9

I definitely think that it is still really early, but ideas and writing also diffused a lot more slowly in the the 19th century...https://t.co/0rdR3MpYso

Introducing Glitch. Anyone can become a streamer - create rooms, and let the audience decide what happens next. Built with @fal + @grok + @Hailuo_AI . These tools are absolute fire. 🔥 Create your first room ↓ https://t.co/a5kJjc9if0

today we're launching Frontier Computing (YC S26) we grow neurons that colocate memory and compute, so you can more easily train ML models. here's one of ours playing Frogger. and we're building a 500M-neuron cluster. live EOY 2026. that's 2,500x larger than the major biocompute systems available today and we'll keep scaling until large model training on biocompute is widely accessible. after all, human intelligence is built on a lipid, amino acid, glucose and vitamin cofactor soup: eating food is less costly than eating GPU compute

Introducing CUA-Lite 🧵 — an open platform for computer-use agents. Training and benchmarking CUAs (Computer-Use Agents) requires four core pieces: 1️⃣ Agents — the models and the scaffolding that drives them 2️⃣ Environments — runtime/sandboxes for agents to interact with, tasks & verifiers/graders 3️⃣ Traces — records of agent trajectories 4️⃣ Frameworks — to evaluate, SFT & RL-train agents Today, all four are fragmented. Every agent ships with its own implementation, often in a separate repo — there is no unified way to run them all. Every environment exposes its own interface and action space, often requiring an expensive VM sandbox for each verifiable task. Traces come in incompatible formats. And without common standards across the stack, every project ends up rebuilding its own tooling/framework for eval, SFT, and RL. CUA-Lite unifies the stack: → One standardized interface & action space for agents and environments → One standardized format for agent traces → One framework for evaluation, SFT & RL → Across desktop, browser & mobile And open resources plug straight in, creating the largest open collection of CUA agents, environments and traces, all in a unified format: 🤖 10+ CUAs, including GPT, Claude, Gemini, Qwen, Muse-Glimmer, UI-TARS 🌐 15+ benchmarks, including OSWorld, WebArena & AndroidWorld ⚡ Optional VM-free sandboxes with 30K+ verifiable tasks for training 📚 10+ trace datasets, freely available on Hugging Face, including public datasets converted into the standardized format and fresh rollouts from frontier open-weight CUAs Led by @BerkeleyRDI , our goal is for CUA-Lite to become a community-driven, open-source ecosystem for computer-use agents. Join the community and contribute today: bring an environment (runtime/sandbox + tasks + verifier), traces, or an agent, and plug it into CUA-Lite!

🚀 Join us for MCP Live on September 9! Hear from experts across the MCP ecosystem and explore what’s new with MCP. Register today: https://t.co/i9thDysTgd https://t.co/qs5sH50PLf

@okzdm_ https://t.co/WjOuJgtVIn

@Vintage5tyle https://t.co/WjOuJgtVIn

@tenkafireNo1 https://t.co/WjOuJgtVIn

@aikotoba_a_o https://t.co/WjOuJgtVIn